مدل ترکیبی شبکه های عصبی و الگوریتم ژنتیک برای طراحی کنترل در سیستم های مبتنی بر اینترنت در تجارت الکترونیکB to C

As organizations become increasingly dependent on Internet-based systems for business-to-consumer electronic commerce (ISB2C), the issue of IS security becomes increasingly important. As the usage of security controls is related to the implementation of ISB2C, the extent of ISB2C controls can be adjusted in order to enable the greatest extent of implementation of ISB2C. This study intends to propose ISB2C-NNGA (ISB2C-controls design using neural networks and genetic algorithms), a hybrid optimization model using neural networks and genetic algorithms for the design of ISB2C controls, which uses back-propagation neural networks (BPN) model as a prediction of controls using system environments, and GA as a pattern directed search mechanism to estimate the exponent of independent variables (i.e., ISB2C controls) in multivariate regression analysis of power model. The effect of system environments on controls can be estimated using BPN model which outperformed linear regression analysis in terms of square root of mean squared error. The effect of each mode of controls on implementation (volume) can be identified using exponents and standardized coefficients in the GA-based nonlinear regression analysis in ISB2C-NNGA. ISB2C-NNGA outperformed conventional linear regression analysis in prediction accuracy in terms of the average R square and sum of squared error. ISB2C can suggest the best set of values for controls to be recommended from several candidate sets of values for controls by identifying the set of values for controls which produce greatest extent of ISB2C implementation. The results of study will support the design of ISB2C controls effectively.

The Internet has changed the way people think, do business, and communicate with each other. Internet-based systems for business-to-consumer electronic commerce (ISB2C) is an application of electronic commerce that allows firms conduct business transactions with consumers over Internet-based information systems such as shopping malls, portals, and web-based systems. The downside to this is that while online, all Internet-based electronic commerce is vulnerable to misuse either by unauthorized users penetrating the system or by authorized users abusing their privileges. The 2008 Computer Security Institute/Federal Bureau of Investigation (CSI/FBI) found that respondents’ estimate of the losses caused by various types of computer security incident was $288,618 for the 522 respondents (Computer Security Institute, 2008). The CSI study indicated that almost half of companies had experienced one to five security incidents in the previous year. Although IS security is not the only one slowing down the proliferation of e-commerce, lack of security is still one of the most likely reason for the low utilization of online selling and electronic payment systems (Lee et al., 1998 and Suh and Han, 2003). In the context of ISB2C, this paper provides a novel approach to selecting and recommending the appropriate controls for successful implementation. The approach provides a back-propagation neural networks and genetic algorithm (GA) based approach, i.e., ISB2C-NNGA (ISB2C controls design using neural networks and genetic algorithms), a hybrid optimization model using neural networks and genetic algorithms for the design of ISB2C controls, which uses back-propagation neural networks (BPN) model as a prediction of controls using system environments, and GA as a pattern directed search mechanism to estimate the exponent of independent variables (i.e., ISB2C controls) in nonlinear regression analysis of power model. The effect of system environments on controls can be estimated using BPN model. The effect of each mode of controls on implementation (volume) can be identified using exponents and standardized coefficients in the GA-based nonlinear regression analysis in ISB2C-NNGA. ISB2C can suggest the best set of values for controls to be recommended from several candidate sets of values for controls by identifying the set of values for controls which produce greatest extent of ISB2C implementation. This approach is a hybrid approach that combines BPN and GA to identify the extent of controls that maximizes the ISB2C implementation. GA based nonlinear regression analysis is used to identify the set of values for controls which lead to greatest ISB2C implementation. In order to evaluate the effectiveness of ISB2C-NNGA, the prediction accuracy of ISB2C-NNGA is compared with that of multivariate linear regression analysis.

This study has proposed the hybrid model for the design of ISB2C controls based on the relationship among system environments, controls, and implementation. In this study, BPN was used for the step of identifying the effect of system environments, i.e., top management support, system compatibility, IS infrastructure, IS expertise, and information contents, on controls, and GA based nonlinear regression model was used for the step of discovering the effect of controls and implementation and the best set of values for controls to be recommended. The proposed model, i.e., BPN and GA-based nonlinear regression model outperformed the conventional linear regression model analysis. From the experimental results, the proposed model has advantages when the model analyzes the data with complex and nonlinear relationships among system environments, controls, and implementation. In the context of ISB2C, the study investigates the effect of five system environments on controls. From BPN model, the effect of system environments on controls (EECji) can be compared among different pairs of system environments and ISB2C controls. For instance, the effect of IS infrastructure on contingency planning controls (C1_1) is greater than the other system environments. This represents the importance of IS infrastructure as a major positive predictor of the use of C1_1. Sophistication is the most critical factor to controls, as the effects of this variable on six (among 10) modes of controls are higher than the other input variables. The GA-based nonlinear regression model recommends the important controls and set of values for controls among candidate sets of values using GA based optimization of regression analysis. In terms of the effect of controls on implementation (ECIi) computed from standardized coefficient and estimated exponents, correction controls for processing integrity (C3_2) is the most important ISB2C controls for both volume of ISB2C, indicating auditors should invest much IT resources for the development and operation of the controls. The GA-based nonlinear regression model suggests the important ISB2C controls for ISB2C implementation. The predicted extent of implementation can be compared with present extent of implementation and if the former is greater than the latter, the present extent of controls represents the appropriate level of controls. Based on the premise that better controls lead to better implementation, this study further suggests that the set of values for controls which has greatest predicted extent of implementation and which most improves the present extent of implementation is the most desirable set of values for controls. IT auditor can place more resources to adjust the extent of each control to follow these values of the optimal set. This shows that the combinations of values of each control can lead to different extent of implementation, and auditor should be careful in adjusting the extent of each control. The results of the study can be used to persuade management who may be reluctant to implement controls because of cultural resistance or decrease of system response time and utilization. The results can help investment of resources to crucial mode of controls and adjustment of the investment on the extent of specific mode of controls.